Spatial Filterative Stochastic Gene Optimized Feature Selection Based Deep Jaccard Regressive Squeezenet Learning For Land Cover Change Detection With Satellite Images

Main Article Content

Mrs A. Jensila Smile , Dr (Mrs) C. Immaculate Mary

Abstract

Change detection is a significant process in satellite surveillance. With the availability of satellite images of a certain geographical area captured in dissimilar time instances, change detection is considered a tough task. Land cover classification of satellite images has been a very biggest area since the amount of information acquired by satellite imaging systems is more. Satellite images reveal spatial and/or temporal information's in which the conventional machine learning algorithms fail to perform well for accurate change detection. To improve the land cover change detection rate, a novel technique called Spatial Filterized Stochastic Gene Optimization-based Deep Jaccard Regressive Squeezenet Convolutional Structure learning (SFSGO-DJRSCSL) is introduced. The SFSGO-DJRSCSL technique consists of three major processes namely preprocessing, feature selection, and change detection. At first, the input satellite images at a different time are captured from the dataset. After collecting the images, the proposed SFSGO-DJRSCSL starts to perform the preprocessing with the input satellite images captured at a different time. After the preprocessing, the stochastic roulette gene optimization based feature selection is carried out to choose the optimal feature combination for change detection resulting in it reduces time consumption. The extracted features are given to the Deep Jaccard Regressive Squeezenet Convolutional Structure learning to detect the changes between homogeneous images. Jaccard index Regression function is applied to map the features extracted from the two homogeneous images and accurately perform the change detection with a lesser error rate. An experimental assessment of the proposed SFSGO-DJRSCSL technique is carried out using a satellite image dataset. The results are discussed with the different performance metrics such as detection rate, false alarm rate, the detection time for different satellite images.

Article Details

Section
Articles